Ceph

Hello, I noticed weird scrubing process. I have a cluster with 3 nodes and 29 OSDs and 1090 pgs. I see that deep-scrubing is very slow and can't scrub all pg in the right time. Now i have 852 pgs not deep-scrubbed in time. Scrubing process is described by charts below:

Why after starts scrubing with 13 pgs it's reduced to 1? I was tried increase osd_max_scrub but peak at the begin caused very high IO and my clients down. Why there is only one scrub almost all the time?

osd_max_scrubs = 3
osd_scrub_begin_hour = 22
osd_scrub_end_hour = 8
osd_scrub_begin_week_day = 0
osd_scrub_end_week_day = 0
osd_scrub_min_interval = 259200 
osd_scrub_max_interval = 604800
osd_deep_scrub_interval = 2419200
osd_deep_scrub_randomize_ratio = 0.02